Leveraging Machine Learning for Threat Detection: A Study on Malicious URL Classification

In today’s digital age, the internet serves as a vital resource. However, it poses significant risks from malicious websites that threaten cybersecurity. This study focuses on developing effective threat detection methods to identify harmful URLs using machine learning models, specifically Random Forest and XGBoost, to classify URL types such as benign, phishing, malware, and defacement. A comprehensive feature extraction process trains the classification models for real-time analysis of URL threats. The findings indicate that the Random Forest model outperforms XGBoost, particularly in classifying phishing URLs, while demonstrating robust accuracy despite challenges with overfitting. This research culminates in the creation of an Android-based Malicious URL Detector mobile application that integrates the trained model, aiming to enhance cybersecurity defenses. By leveraging machine learning techniques, this study contributes to more efficient and effective detection of malicious URLs, fostering a safer online environment for users.